The spine of my research: how groups of agents — human, artificial, or mixed — come to act together, model one another, and concentrate or distribute power.
- A Robust Measure of Collective Power — what agents can achieve alone and together, and the failure modes of maximising it (slides)
- Stewarding Power in Human-AI Collectives — incentive structures and social norms against power concentration
- Beyond Individual Power — information-theoretic bounds on collective power (slides)
- Factorised Active Inference for Strategic Multi-Agent Interactions — agents keeping individual beliefs while planning in a joint context
- Inverse Reinforcement Learning as the Algorithmic Basis for Theory of Mind — review of IRL methods as a computational account of ToM
See also PhD projects for the papers themselves, and ethics-and-political-economy where the question turns normative.